Zoom Screen-Sharing Flaw Cracked With Fewer Than 20 AI Prompts
Vulnerability let attackers remotely control cameras and microphones by abusing the annotation feature, discovered and patched after A Security's find
90
Tag
security
OpenAI unveils GPT-5.6-Cyber, a model dedicated to cybersecurity
Daybreak program split into Blue and Red tiers, with access restricted to approved defenders
20
Anthropic mandates watermarking on Claude outputs worldwide
Under EU AI Act rules, invisible markers for text and signed provenance data for images start August 2026
1240
Apple unveils technique to block fine-tuning of AI model weights
"DLR-Lock" research lets users run open-weight models as-is while blocking modification
20
Cohere CEO: "Open-source models need three things"
Aidan Gomez explains why the company is expanding its open-source lineup and how it's designing for "sovereign" deployment
20
Atlassian AI Agent Breached via White Text Hidden in PDF
Security firm PromptArmor discloses vulnerability letting Rovo exfiltrate Jira and Confluence data without a trace
90
That's the last story.